The DVD (Digital Versatile Disk) format was designed by various members of the consumer electronics industry as a means of storing high quality audio-video content (e.g., a feature length film) on a single disk. To facilitate such efficient storage, the DVD format uses contemporary compression technologies to reduce the sizes of the video and audio bit streams comprising the content.
The DVD format employs the ISO MPEG-2 standard to compress video. MPEG-2 represents video content as a compressed series of frames. Each frame is a rectangular array of picture elements (pixels) depicting the content at a particular instant in time. Thus playback consists of decompressing and then displaying this series of frames.
In conventional DVD players, the playback of DVD content is typically implemented in one of three ways: using dedicated hardware, using a software implementation, or using a combination of software and hardware. The most common or conventional implementation takes the form of a consumer electronics components with limited resources targeted exclusively at DVD playback. Less common or conventional implementations take the form of PC-based implementations or Game Consoles (e.g., Sony PLAYSTATION2 and Microsoft X-BOX) which, because they target multiple functions, tend to have more extensive resources.
Although the user primarily plays DVD content forwards and at the display rate of the original content, there are circumstances where a user may want to display frames either backwards or at a rate other than the normal display rate or both backwards and at a rate other than the normal display rate. These include playback features popularized by VCRs, e.g., slow motion, backwards play, single frame stepping forwards and backwards, as well as fast forward and rewind at various rates. The realization of alternate forms of playback, however, is hindered by both the nature of MPEG-2 compression and of typical DVD players.
The MPEG-2 video standard employs three types of compressed frames: intra-frames (I-frames), predictive frames (P-frames), and bi-directionally predictive frames (B-frames). I-frames have no dependencies. Thus, an I-frame is self-contained and includes all information necessary to reproduce the associated original frame. P-frames may have forward dependencies, i.e., a P-frame is not self-contained. It may re-use information from the preceding decompressed reference frame (where a reference frame is either an I-frame or another P-frame). Thus, a playback implementation must decode reference frame preceding a P-frame before it decodes the P-frame itself and must keep the preceding reference frame resident in memory throughout the decoding of a P-frame. For the B-frames, B-frames are not self-contained and may have forward and backward dependencies, such that it may re-use information from either the preceding or the subsequent decompressed reference frame. Thus, a playback implementation must decode the reference frames both preceding and following a B-frame before it decodes the B-frame itself and must keep both the preceding and subsequent reference frames resident in memory throughout the decoding of a B-frame.
In summary, a playback implementation respects the dependencies between frames by decoding a frame's references before decoding the frame itself. It also keeps the required references in memory until all frames requiring the references are decoded.
Normally, a playback implementation maintains four frame buffers (i.e., arrays of memory) of MPEG-2 video at any one time:
(1) currently decoded frame
(2) forwards reference. The reference preceding the currently decoded frame. Sometimes co-incident with the currently displayed frame.
(3) backwards reference. The reference following the currently decoded frame.
(4) currently displayed frame. This is distinct from the currently decoded frame to prevent the player from updating a frame while it is being displayed which causes an undesirable visual artifact called “tearing”.
The maintenance of only four or five frame buffers helps to limit the memory resources used in a DVD player, while providing sufficient forwards playback. Four (or even five) frames, however, is normally not sufficient for smooth backwards playback of MPEG-2. Rather, DVD players currently avoid decoding P or B frames when playing backwards, and only decode the self-contained I-frames. This leads to decoding less than 10% of the frames and subsequently to jerky, low frame-rate, backwards play. Further, decoding only the I-frames prevents DVD players from single stepping frames backwards. While an implementation could repeat a process of starting from the last I-frame and progressing forward using only four frames, the process is laborious, requiring repetition for every new backwards frame at the normal playback speed, which requires decoding capacity far in excess of even current PC-based implementations.
Furthermore, most DVD players output to televisions or other displays with fixed display rates. For instance, the PAL and NTSC television standards mandate display rates of 25 and 29.97 frames per second. This is sufficient for normal speed playback which is limited, by the DVD standard, to PAL and NTSC rates. This limitation, however, prevents higher frame rate display during fast forward or rewind. For instance, if the player produces 60 frames of video per second in a 2× fast forward mode only 30 (i.e., every other frame decoded) may be displayed on an NTSC television. Further, given fixed frame rates, these players cannot smoothly vary the rate in a transition from one rate (say normal speed) to another (say 2× speed).
Accordingly, what is needed are a system and method for smooth rewind, single frame stepping backwards, and smooth search transitions in a DVD implementation. The present invention addresses such a need.